doubt about featurizer tensor size

[pmnerfed]

Hi,
According to the code, the tensor created is tensor[0][init_idx][attr_idx] = 1.0.
Can you please elaborate on why its not tensor[0][attr_idx][init_idx] = 1.0(axis 1 and 2 swapped).

file : repair/featurize/initattfeat.py@12
From what I understand,
if each row in the tensor should be for an attribute, it should be the axis=1 not 2.
The axis =2 should be for valid value for the attribute the row corresponds to.

Hope that makes sense.
If my thinking is wrong, could you please help me out here?
`

[fatangare]

I believe it is because all featurized tensors are concatenated in later stage. If you have diff dimensions for initattfeat feature , tensors can't be concatenated.